Discuss Findings

all-sanitized follow-up (1733 results, 51 hosts, host-HHI=0.03792, top-host share=8.25%) shows best_mid_gain_mean 0.39599 +/- 0.00015 (95% CI 0.39570-0.39629, d=63.57532, sign-consistent=100.0%, median 0.39620, q05-q95 0.38604-0.40511, skew=-0.34); memory_synergy_mean 0.39301 +/- 0.00015 (95% CI 0.39272-0.39331, d=62.45603, sign-consistent=100.0%); positive_mid_gain_fraction_mean 0.99311 +/- 8.60e-05 (95% CI 0.99294-0.99328); recovery_advantage_mean -0.02006 +/- 2.74e-05 (95% CI -0.02012 to -0.02001, d=-17.57361, sign-consistent=0.0%); mean_diversity_advantage -0.17089 +/- 3.87e-05 (95% CI -0.17097 to -0.17082, d=-106.05706, sign-consistent=0.0%); cat=6 terminal_coexistence averaged over inheritance/asymmetry is memory 0.0: 0.00298 +/- 0.00048, memory 0.55: 0.63457 +/- 0.10202, memory 0.9: 0.94465 +/- 0.00505, with mid-memory gains vs its no-cat baseline ranging 0.32846-0.44704 across inheritance/asymmetry and the strongest absolute mid-memory cat=6 coexistence at inheritance 0.35, asymmetry 0.0-0.35 (~0.864).

CONFIRMED, with a mechanistic qualification. Across the full sanitized family, moderate catastrophe pulses robustly rescue the intermediate-memory regime relative to both its no-catastrophe baseline and the zero-memory controls, so the catastrophe-plus-lineage-memory interaction is real and very stable. The likely mechanism is that nonzero founder-memory prevents complete post-bottleneck washout, letting exchange reseed depleted compartments while keeping parasite burden extremely low, which matches the large positive synergy term and the jump from ~0.239 to ~0.635 terminal coexistence at memory 0.55 when cat=6 is introduced. However, the deeper aggregate does not support the stronger claim that near-perfect memory blocks rescue: memory 0.9 remains even more coexistence-preserving and more diverse in absolute terms, while both recovery-advantage and diversity-advantage contrasts are stably negative. The implication is that lineage memory clearly helps catastrophe resilience, but the family now looks more like a strong monotone or saturating memory benefit than a strict intermediate optimum.